Re: PR/47404 CVS commit: pkgsrc/devel/py-pysvn

[Richard PALO <richard%netbsd.org@localhost>]

Le 20/05/13 07:40, Richard PALO a écrit :
> The following reply was made to PR pkg/47404; it has been noted by GNATS.
> 
> From: Richard PALO <Richard.PALO%baou.fr@localhost>
> To: gnats-bugs%NetBSD.org@localhost
> Cc: 
> Subject: Re: PR/47404 CVS commit: pkgsrc/devel/py-pysvn
> Date: Mon, 20 May 2013 07:39:51 +0200
> 
>  Le 20/04/13 19:25, Richard PALO a écrit :
>  > The following reply was made to PR pkg/47404; it has been noted by GNATS.
>  >
>  > From: Richard PALO <Richard.PALO%baou.fr@localhost>
>  > To: gnats-bugs%NetBSD.org@localhost
>  > Cc:
>  > Subject: Re: PR/47404 CVS commit: pkgsrc/devel/py-pysvn
>  > Date: Sat, 20 Apr 2013 19:20:31 +0200
>  >
>  >   This is a multi-part message in MIME format.
>  >   --------------060201000800000203070908
>  >   Content-Type: text/plain; charset=UTF-8; format=flowed
>  >   Content-Transfer-Encoding: 8bit
>  >
>  >   Le 15/04/13 15:35, OBATA Akio a écrit :
>  >   > The following reply was made to PR pkg/47404; it has been noted by GNATS.
>  >   >
>  >   > From: "OBATA Akio" <obache%netbsd.org@localhost>
>  >   >   To generate a diff of this commit:
>  >   >   cvs rdiff -u -r1.18 -r1.19 pkgsrc/devel/py-pysvn/Makefile
>  >   >
>  >
>  >   First, testing *without* the above patch shows that the latest
>  >   (upgraded) version of py-pysvn seems to build fine.
>  >
>  >   Therefore I think this patch is ultimately unnecessary.
>  >
>  >   As Mr Holland suggested, I took a look at "fixing" the wrappers...
>  >
>  >   First, the following:
>  >   > richard@x3200:~/src/pkgsrc/mk/wrapper$ git diff master .
>  >   > diff --git a/mk/wrapper/arg-source b/mk/wrapper/arg-source
>  >   > index 111bb4d..f6231a5 100644
>  >   > --- a/mk/wrapper/arg-source
>  >   > +++ b/mk/wrapper/arg-source
>  >   > @@ -88,6 +88,7 @@ while $test $# -gt 0; do
>  >   >         ##############################################################
>  >   >         # Split "-Wl,-R/dir1:/dir2" into "-Wl,-R/dir1 -Wl,-R/dir2".
>  >   >         # Same for -R and -Wl,-rpath and -Wl,-rpath-link.
>  >   > +       # (at the same time suppress the double dash in --rpath to -rpath
>  >   >         ##############################################################
>  >   >         -R*:*|-Wl,-R*:*|\
>  >   >         -Wl,-rpath,*:*|-Wl,-rpath-link,*:*|-Wl,--rpath,*:*)
>  >   > @@ -96,7 +97,7 @@ while $test $# -gt 0; do
>  >   >                 -Wl,-R*)                R="-Wl,-R" ;;
>  >   >                 -Wl,-rpath,*)           R="-Wl,-rpath," ;;
>  >   >                 -Wl,-rpath-link,*)      R="-Wl,-rpath-link," ;;
>  >   > -               -Wl,--rpath,*)          R="-Wl,--rpath," ;;
>  >   > +               -Wl,--rpath,*)          R="-Wl,-rpath," ;;
>  >   >                 esac
>  >   >                 list="${arg#$R}"
>  >   >                 save_IFS="${IFS}"; IFS=":"
>  >   > @@ -111,6 +112,7 @@ while $test $# -gt 0; do
>  >   >         # Merge and split "-Wl,-R -Wl,/dir1:/dir2" into
>  >   >         # "-Wl,-R/dir1 -Wl,-R/dir2".  Same for -Wl,-rpath and
>  >   >         # -Wl,-rpath-link.
>  >   > +       # (at the same time suppress the double dash in --rpath to -rpath
>  >   >         ##############################################################
>  >   >         -Wl,-R|-Wl,-rpath|-Wl,-rpath-link|-Wl,--rpath)
>  >   >                 nextarg="$1"; shift
>  >   > @@ -118,7 +120,7 @@ while $test $# -gt 0; do
>  >   >                 -Wl,-R)                 R="-Wl,-R" ;;
>  >   >                 -Wl,-rpath)             R="-Wl,-rpath," ;;
>  >   >                 -Wl,-rpath-link)        R="-Wl,-rpath-link," ;;
>  >   > -               -Wl,--rpath)            R="-Wl,--rpath," ;;
>  >   > +               -Wl,--rpath)            R="-Wl,-rpath," ;;
>  >   >                 esac
>  >   >                 nextarg="${nextarg#-Wl,}"
>  >   >                 case $nextarg in
>  >
>  >   seems to be able to do the trick...
>  >
>  >   I originally worked on transform-gcc but this seemed to be a more global
>  >   solution.
>  >
>  >   At the moment, in any pkgsrc/*/Makefile I can only easily find
>  >   pkgsrc/chat/inspircd which also has a workaround put in for this issue
>  >   (but seems a bit broken since migration to openssl 1.0.1x, at least on
>  >   solaris)...
>  >
>  >   By searching patches, I notice no real workarounds, but that there do
>  >   exist a number of cases of '--rpath' used and as well '--rpath='
>  >
>  >   I believe the split case above could be adapted to easily handle
>  >   '--rpath=' in addition to '--rpath,' .
>  >
>  >   This seems in any event a reasonable patch for the case cited originally
>  >   *as*, after searching _COMPILER_RPATH_FLAG and _LINKER_RPATH_FLAG (nor
>  >   _OPSYS_COMPILER_RPATH_FLAG), I can find no explicit instances of "--rpath".
>  >
>  >   Observations?
>  >
>  I don't have a problem with the "quick" fix to py-pysvn,
>  but perhaps consideration should be taken for a possible global solution 
>  as Mr Holland suggested perhaps with an adaptation of the above patch if 
>  anybody else has tested it...
>  
>  -- 
>  Richard PALO

This bug report is still open (even though the quick fix to py-pysvn was made by obache@ at the time),
I'm still in favour of fixing mk/wrapper/arg-source after the freeze is lifted.

Any objections?

-- 
Richard PALO